Merge pull request #6741 from esphome/bump-2024.5.0b5

2024.5.0b5

esphome/components/api/api.proto

@@ -1147,6 +1147,9 @@ message MediaPlayerCommandRequest {
 
   bool has_media_url = 6;
   string media_url = 7;
+
+  bool has_announcement = 8;
+  bool announcement = 9;
 }
 
 // ==================== BLUETOOTH ====================

esphome/components/api/api_connection.cpp

@@ -1002,7 +1002,11 @@ bool APIConnection::send_media_player_state(media_player::MediaPlayer *media_pla
 
   MediaPlayerStateResponse resp{};
   resp.key = media_player->get_object_id_hash();
-  resp.state = static_cast<enums::MediaPlayerState>(media_player->state);
+
+  media_player::MediaPlayerState report_state = media_player->state == media_player::MEDIA_PLAYER_STATE_ANNOUNCING
+                                                    ? media_player::MEDIA_PLAYER_STATE_PLAYING
+                                                    : media_player->state;
+  resp.state = static_cast<enums::MediaPlayerState>(report_state);
   resp.volume = media_player->volume;
   resp.muted = media_player->is_muted();
   return this->send_media_player_state_response(resp);
@@ -1038,6 +1042,9 @@ void APIConnection::media_player_command(const MediaPlayerCommandRequest &msg) {
   if (msg.has_media_url) {
     call.set_media_url(msg.media_url);
   }
+  if (msg.has_announcement) {
+    call.set_announcement(msg.announcement);
+  }
   call.perform();
 }
 #endif

esphome/components/api/api_pb2.cpp

@@ -5253,6 +5253,14 @@ bool MediaPlayerCommandRequest::decode_varint(uint32_t field_id, ProtoVarInt val
       this->has_media_url = value.as_bool();
       return true;
     }
+    case 8: {
+      this->has_announcement = value.as_bool();
+      return true;
+    }
+    case 9: {
+      this->announcement = value.as_bool();
+      return true;
+    }
     default:
       return false;
   }
@@ -5289,6 +5297,8 @@ void MediaPlayerCommandRequest::encode(ProtoWriteBuffer buffer) const {
   buffer.encode_float(5, this->volume);
   buffer.encode_bool(6, this->has_media_url);
   buffer.encode_string(7, this->media_url);
+  buffer.encode_bool(8, this->has_announcement);
+  buffer.encode_bool(9, this->announcement);
 }
 #ifdef HAS_PROTO_MESSAGE_DUMP
 void MediaPlayerCommandRequest::dump_to(std::string &out) const {
@@ -5323,6 +5333,14 @@ void MediaPlayerCommandRequest::dump_to(std::string &out) const {
   out.append("  media_url: ");
   out.append("'").append(this->media_url).append("'");
   out.append("\n");
+
+  out.append("  has_announcement: ");
+  out.append(YESNO(this->has_announcement));
+  out.append("\n");
+
+  out.append("  announcement: ");
+  out.append(YESNO(this->announcement));
+  out.append("\n");
   out.append("}");
 }
 #endif

esphome/components/api/api_pb2.h

@@ -1298,6 +1298,8 @@ class MediaPlayerCommandRequest : public ProtoMessage {
   float volume{0.0f};
   bool has_media_url{false};
   std::string media_url{};
+  bool has_announcement{false};
+  bool announcement{false};
   void encode(ProtoWriteBuffer buffer) const override;
 #ifdef HAS_PROTO_MESSAGE_DUMP
   void dump_to(std::string &out) const override;

esphome/components/i2s_audio/media_player/i2s_audio_media_player.cpp

@@ -10,14 +10,19 @@ namespace i2s_audio {
 static const char *const TAG = "audio";
 
 void I2SAudioMediaPlayer::control(const media_player::MediaPlayerCall &call) {
+  media_player::MediaPlayerState play_state = media_player::MEDIA_PLAYER_STATE_PLAYING;
+  if (call.get_announcement().has_value()) {
+    play_state = call.get_announcement().value() ? media_player::MEDIA_PLAYER_STATE_ANNOUNCING
+                                                 : media_player::MEDIA_PLAYER_STATE_PLAYING;
+  }
   if (call.get_media_url().has_value()) {
     this->current_url_ = call.get_media_url();
     if (this->i2s_state_ != I2S_STATE_STOPPED && this->audio_ != nullptr) {
       if (this->audio_->isRunning()) {
         this->audio_->stopSong();
       }
       this->audio_->connecttohost(this->current_url_.value().c_str());
-      this->state = media_player::MEDIA_PLAYER_STATE_PLAYING;
+      this->state = play_state;
     } else {
       this->start();
     }
@@ -35,7 +40,7 @@ void I2SAudioMediaPlayer::control(const media_player::MediaPlayerCall &call) {
       case media_player::MEDIA_PLAYER_COMMAND_PLAY:
         if (!this->audio_->isRunning())
           this->audio_->pauseResume();
-        this->state = media_player::MEDIA_PLAYER_STATE_PLAYING;
+        this->state = play_state;
         break;
       case media_player::MEDIA_PLAYER_COMMAND_PAUSE:
         if (this->audio_->isRunning())
@@ -126,7 +131,9 @@ void I2SAudioMediaPlayer::loop() {
 
 void I2SAudioMediaPlayer::play_() {
   this->audio_->loop();
-  if (this->state == media_player::MEDIA_PLAYER_STATE_PLAYING && !this->audio_->isRunning()) {
+  if ((this->state == media_player::MEDIA_PLAYER_STATE_PLAYING ||
+       this->state == media_player::MEDIA_PLAYER_STATE_ANNOUNCING) &&
+      !this->audio_->isRunning()) {
     this->stop();
   }
 }
@@ -164,6 +171,10 @@ void I2SAudioMediaPlayer::start_() {
   if (this->current_url_.has_value()) {
     this->audio_->connecttohost(this->current_url_.value().c_str());
     this->state = media_player::MEDIA_PLAYER_STATE_PLAYING;
+    if (this->is_announcement_.has_value()) {
+      this->state = this->is_announcement_.value() ? media_player::MEDIA_PLAYER_STATE_ANNOUNCING
+                                                   : media_player::MEDIA_PLAYER_STATE_PLAYING;
+    }
     this->publish_state();
   }
 }

esphome/components/i2s_audio/media_player/i2s_audio_media_player.h

@@ -78,6 +78,7 @@ class I2SAudioMediaPlayer : public Component, public media_player::MediaPlayer,
   HighFrequencyLoopRequester high_freq_;
 
   optional<std::string> current_url_{};
+  optional<bool> is_announcement_{};
 };
 
 }  // namespace i2s_audio

esphome/components/media_player/__init__.py

@@ -51,12 +51,16 @@
 
 CONF_ON_PLAY = "on_play"
 CONF_ON_PAUSE = "on_pause"
+CONF_ON_ANNOUNCEMENT = "on_announcement"
 CONF_MEDIA_URL = "media_url"
 
 StateTrigger = media_player_ns.class_("StateTrigger", automation.Trigger.template())
 IdleTrigger = media_player_ns.class_("IdleTrigger", automation.Trigger.template())
 PlayTrigger = media_player_ns.class_("PlayTrigger", automation.Trigger.template())
 PauseTrigger = media_player_ns.class_("PauseTrigger", automation.Trigger.template())
+AnnoucementTrigger = media_player_ns.class_(
+    "AnnouncementTrigger", automation.Trigger.template()
+)
 IsIdleCondition = media_player_ns.class_("IsIdleCondition", automation.Condition)
 IsPlayingCondition = media_player_ns.class_("IsPlayingCondition", automation.Condition)
 
@@ -75,6 +79,9 @@ async def setup_media_player_core_(var, config):
     for conf in config.get(CONF_ON_PAUSE, []):
         trigger = cg.new_Pvariable(conf[CONF_TRIGGER_ID], var)
         await automation.build_automation(trigger, [], conf)
+    for conf in config.get(CONF_ON_ANNOUNCEMENT, []):
+        trigger = cg.new_Pvariable(conf[CONF_TRIGGER_ID], var)
+        await automation.build_automation(trigger, [], conf)
 
 
 async def register_media_player(var, config):
@@ -106,6 +113,11 @@ async def register_media_player(var, config):
                 cv.GenerateID(CONF_TRIGGER_ID): cv.declare_id(PauseTrigger),
             }
         ),
+        cv.Optional(CONF_ON_ANNOUNCEMENT): automation.validate_automation(
+            {
+                cv.GenerateID(CONF_TRIGGER_ID): cv.declare_id(AnnoucementTrigger),
+            }
+        ),
     }
 )
 

esphome/components/media_player/automation.h

@@ -52,6 +52,7 @@ class StateTrigger : public Trigger<> {
 MEDIA_PLAYER_SIMPLE_STATE_TRIGGER(IdleTrigger, IDLE)
 MEDIA_PLAYER_SIMPLE_STATE_TRIGGER(PlayTrigger, PLAYING)
 MEDIA_PLAYER_SIMPLE_STATE_TRIGGER(PauseTrigger, PAUSED)
+MEDIA_PLAYER_SIMPLE_STATE_TRIGGER(AnnouncementTrigger, ANNOUNCING)
 
 template<typename... Ts> class IsIdleCondition : public Condition<Ts...>, public Parented<MediaPlayer> {
  public:

esphome/components/media_player/media_player.cpp

@@ -15,6 +15,8 @@ const char *media_player_state_to_string(MediaPlayerState state) {
       return "PLAYING";
     case MEDIA_PLAYER_STATE_PAUSED:
       return "PAUSED";
+    case MEDIA_PLAYER_STATE_ANNOUNCING:
+      return "ANNOUNCING";
     case MEDIA_PLAYER_STATE_NONE:
     default:
       return "UNKNOWN";
@@ -68,6 +70,9 @@ void MediaPlayerCall::perform() {
   if (this->volume_.has_value()) {
     ESP_LOGD(TAG, "  Volume: %.2f", this->volume_.value());
   }
+  if (this->announcement_.has_value()) {
+    ESP_LOGD(TAG, " Announcement: %s", this->announcement_.value() ? "yes" : "no");
+  }
   this->parent_->control(*this);
 }
 
@@ -108,6 +113,11 @@ MediaPlayerCall &MediaPlayerCall::set_volume(float volume) {
   return *this;
 }
 
+MediaPlayerCall &MediaPlayerCall::set_announcement(bool announce) {
+  this->announcement_ = announce;
+  return *this;
+}
+
 void MediaPlayer::add_on_state_callback(std::function<void()> &&callback) {
   this->state_callback_.add(std::move(callback));
 }

esphome/components/media_player/media_player.h

@@ -10,7 +10,8 @@ enum MediaPlayerState : uint8_t {
   MEDIA_PLAYER_STATE_NONE = 0,
   MEDIA_PLAYER_STATE_IDLE = 1,
   MEDIA_PLAYER_STATE_PLAYING = 2,
-  MEDIA_PLAYER_STATE_PAUSED = 3
+  MEDIA_PLAYER_STATE_PAUSED = 3,
+  MEDIA_PLAYER_STATE_ANNOUNCING = 4
 };
 const char *media_player_state_to_string(MediaPlayerState state);
 
@@ -51,19 +52,22 @@ class MediaPlayerCall {
   MediaPlayerCall &set_media_url(const std::string &url);
 
   MediaPlayerCall &set_volume(float volume);
+  MediaPlayerCall &set_announcement(bool announce);
 
   void perform();
 
   const optional<MediaPlayerCommand> &get_command() const { return command_; }
   const optional<std::string> &get_media_url() const { return media_url_; }
   const optional<float> &get_volume() const { return volume_; }
+  const optional<bool> &get_announcement() const { return announcement_; }
 
  protected:
   void validate_();
   MediaPlayer *const parent_;
   optional<MediaPlayerCommand> command_;
   optional<std::string> media_url_;
   optional<float> volume_;
+  optional<bool> announcement_;
 };
 
 class MediaPlayer : public EntityBase {

esphome/components/voice_assistant/voice_assistant.cpp

@@ -152,7 +152,7 @@ void VoiceAssistant::loop() {
         } else
 #endif
         {
-          this->set_state_(State::START_PIPELINE, State::START_MICROPHONE);
+          this->set_state_(State::START_MICROPHONE, State::START_PIPELINE);
         }
       } else {
         this->high_freq_.stop();
@@ -318,7 +318,7 @@ void VoiceAssistant::loop() {
 #endif
 #ifdef USE_MEDIA_PLAYER
       if (this->media_player_ != nullptr) {
-        playing = (this->media_player_->state == media_player::MediaPlayerState::MEDIA_PLAYER_STATE_PLAYING);
+        playing = (this->media_player_->state == media_player::MediaPlayerState::MEDIA_PLAYER_STATE_ANNOUNCING);
       }
 #endif
       if (playing) {
@@ -514,7 +514,7 @@ void VoiceAssistant::request_start(bool continuous, bool silence_detection) {
     } else
 #endif
     {
-      this->set_state_(State::START_PIPELINE, State::START_MICROPHONE);
+      this->set_state_(State::START_MICROPHONE, State::START_PIPELINE);
     }
   }
 }
@@ -640,7 +640,7 @@ void VoiceAssistant::on_event(const api::VoiceAssistantEventResponse &msg) {
       this->defer([this, url]() {
 #ifdef USE_MEDIA_PLAYER
         if (this->media_player_ != nullptr) {
-          this->media_player_->make_call().set_media_url(url).perform();
+          this->media_player_->make_call().set_media_url(url).set_announcement(true).perform();
         }
 #endif
         this->tts_end_trigger_->trigger(url);

esphome/const.py

@@ -1,6 +1,6 @@
 """Constants used by esphome."""
 
-__version__ = "2024.5.0b4"
+__version__ = "2024.5.0b5"
 
 ALLOWED_NAME_CHARS = "abcdefghijklmnopqrstuvwxyz0123456789-_"
 VALID_SUBSTITUTIONS_CHARACTERS = (